In an operational machine translation system, a variety of texts will be encountered even if its domain of usefulness is restricted to a specific field. This diversity of texts poses a problem on handling word sense ambiguity and customized translation. This paper presents a unification-based method for lexicography that can lessen this problem by dividing the system dictionaries into hierarchically structured general, technical, and customer dictionaries, and also by unifying lexical information in these different dictionaries. In the paper, detailed discussion and example of the unification technique are given. Other advantages of using unification are also noted: the time spent in dictionary construction is saved; dictionary storage space is minimized; the integrity of distinct dictionaries is preserved; customized vocabulary is provided without resorting to run-time dictionary; and the option regarding which dictionaries to be unified is kept open. Besides, in view of the fact that categorial ambiguity might occur as a result of unification, a score function is added as a solution. With these advantages, the unification approach to lexicography is regarded as viable in enhancing the translation performance of a practical machine translation system.